基于PCA和SVM的高光谱遥感图像分类研究

支持向量机(SVM)是根据统计学习理论提出的新的研究方法,它在解决小样本、非线性及高维模式识别问题中表现出了许多特有的优势,在模式识别、函数逼近和概率密度估计等方面取得了良好的效果。由于高光谱图像波段数目多,各波段间具有较强的相关性,因此通过主成分分析(PCA)方法对高光谱数据进行预处理,达到了降维的目的,同时也去除了噪声波段。用支持向量机方法对高光谱遥感图像进行分类,可实现图像的分类识别。     

Rapid Identification of Bacterial Species by Fluorescence Spectroscopy and Classification Through Principal Components Analysis

This work presents the development of a method for rapid bacterial identification based on the autofluorescence spectrum. It was demonstrated differences in the autofluorescence spectrum in three bacterial species and the subsequent separation, through the Principal Components Analysis (PCA) technique, in groups with high likeness, that could identify the bacteria in less than 10 min. Fluorescence spectra of 60 samples of 3 different bacterial species (Escherichia coli, EC, Enterococcus faecalis, EF and Staphylococcus aureus, SA), previously identified by automated equipment Mini API, were collected in 10 excitation wavelengths from 330 to 510 nm. The PCA technique applied to the fluorescence spectra showed that bacteria species could be identified with sensitivity and specificity higher than 90% according to differences that occur within the spectra with excitation of 410 nm and 430 nm. This work presented a method of bacterial identification of three more frequent and more clinically significant species based on the autofluorescence spectra in the excitation wavelengths of 410 and 430 nm and the classification of the spectra in three groups using PCA. The results demonstrated that the bacterial identification is very efficient with such methodology. The proposed method is rapid, ease to perform and low cost compared to standard methods.     

Damage localization in linear-form structures based on sensitivity investigation for principal component analysis

This paper addresses the problem of damage detection and localization in linear-form structures. Principal component analysis (PCA) is a popular technique for dynamic system investigation. The aim of the paper is to present a damage diagnosis method based on sensitivities of PCA results in the frequency domain. Starting from frequency response functions (FRFs) measured at different locations on the structure; PCA is performed to determine the main features of the signals. Sensitivities of principal directions obtained from PCA to structural parameters are then computed and inspected according to the location of sensors; their variation from the healthy state to the damaged state indicates damage locations. It is worth noting that damage localization is performed without the need of modal identification. Influences of some features as noise, choice of parameter and number of sensors are discussed. The efficiency and limitations of the proposed method are illustrated using numerical and real-world examples.     

Structural damage detection of controlled building structures using frequency response functions

If a building structure requires both a vibration control system and a health monitoring system, the integration of the two systems will be cost-effective and beneficial. One of the key problems of this integrated system is how to use control devices to enhance system identification and damage detection. This paper presents a new method for system identification and damage detection of controlled building structures equipped with semi-active friction dampers through model updating based on frequency response functions. The two states of the building are first created by adding a known stiffness using semi-active friction dampers. A scheme based on the frequency response functions of the two states of the building is then presented to identify stiffness parameters of structural members in consideration of structural connectivity and transformation information. By applying the proposed model updating scheme to the damaged building, a damage detection scheme is proposed based on the identified stiffness parameters of structural members of both the original and damaged buildings. The feasibility of the proposed schemes is finally demonstrated through a detailed numerical investigation in terms of an example building, in which the effects of measurement noise and excitation conditions are discussed. The numerical results clearly show that the proposed method can locate and quantify damage satisfactorily even though measurement noise is taken into consideration.     

A damage identification technique based on embedded sensitivity analysis and optimization processes

A vibration based structural damage identification method, using embedded sensitivity functions and optimization algorithms, is discussed in this work. The embedded sensitivity technique requires only measured or calculated frequency response functions to obtain the sensitivity of system responses to each component parameter. Therefore, this sensitivity analysis technique can be effectively used for the damage identification process. Optimization techniques are used to minimize the difference between the measured frequency response functions of the damaged structure and those calculated from the baseline system using embedded sensitivity functions. The amount of damage can be quantified directly in engineering units as changes in stiffness, damping, or mass. Various factors in the optimization process and structural dynamics are studied to enhance the performance and robustness of the damage identification process. This study shows that the proposed technique can improve the accuracy of damage identification with less than 2 percent error of estimation.     

EXPERIMENTAL VALIDATION OF A STRUCTURAL HEALTH MONITORING METHODOLOGY: PART III. DAMAGE LOCATION ON AN AIRCRAFT WING

This paper documents the third phase of a programme of experimental work which validates a structural health monitoring methodology based on novelty detection. In this paper, an extension of the detection method for damage location is proposed and demonstrated. The structure of interest is a Gnat aircraft wing. Although it was not possible to damage the aircraft, the method was demonstrated by determining which of a set of inspection panels had been removed.     

Rolling bearing diagnosing method based on Empirical Mode Decomposition of machine vibration signal

Rolling bearing faults are one of the major reasons for breakdown of industrial machinery and bearing diagnosing is one of the most important topics in machine condition monitoring.The main problem in industrial application of bearing vibration diagnostics is the masking of informative bearing signal by machine noise. The vibration signal of the rolling bearing is often covered or concealed by other structural vibrations sources, such as gears. Although a number of vibration diagnostic techniques have been developed over the last several years, in many cases these methods are quite complicated in use or only effective at later stages of damage development. This paper presents an EMD-based rolling bearing diagnosing method that shows potential for bearing damage detection at a much earlier stage of damage development.By using EMD a raw vibration signal is decomposed into a number of Intrinsic Mode Functions (IMFs). Then, a new method of IMFs aggregation into three Combined Mode Functions (CMFs) is applied and finally the vibration signal is divided into three parts of signal: noise-only part, signal-only part and trend-only part. To further bearing fault-related feature extraction from resultant signals, the spectral analysis of the empirically determined local amplitude is used. To validate the proposed method, raw vibration signals generated by complex mechanical systems employed in the industry (driving units of belt conveyors), including normal and fault bearing vibration data, are used in two case studies. The results show that the proposed rolling bearing diagnosing method can identify bearing faults at early stages of their development.     

Modal analysis by free vibration response only for discrete and continuous systems

This work aims to develop the algorithm for modal analysis by free vibration response only (MAFVRO), in particular for the general or non-proportional viscous damping system model. If the structural displacement or acceleration response due to free vibration can be measured, the system response matrices, including the displacement, velocity and acceleration, can be obtained through numerical differential or integration methods. These response matrices can then be applied to the developed MAFVRO method to determine the structural modal parameters. The numerical differential and integration methods are introduced and adopted to establish the modal parameter prediction program for the non-proportional damping model of MAFVRO. This work also shows the applications of MAFVRO to the multiple degree-of-freedom (mdof) systems and the cantilever beam, respectively. Both the discrete and continuous systems are demonstrated for the feasibility of the MAFVRO algorithm. The developed method uses the free vibration output response only and can obtain the structural modal parameters successfully.     

Frequency response design of uncertain systems using performance indices and meta-models

Systems that are operated near their resonance frequencies experience vibrations that can lead to impaired performance, overstressing, fatigue fracture and adverse human reactions. Frequency response (FR) analysis can be invoked to mitigate the effects. When components of a system are described by random variables, modal frequencies and modal shapes, or, amplitudes and phases, are also random variables and the frequency response (FR) design of the system becomes more complex since it requires the solution of a frequency-variant probability problem. This paper presents a methodology to provide the frequency response design of uncertain systems using a transfer function approach. The methodology is found to be robust, expandable and flexible and can be applied to multi-disciplinary systems with n-dof and multiple design constraints. The novelty of the approach is the creation of a frequency-invariant probability problem through: (a) the discretization of the frequency band of interest into multiple contiguous point frequencies, (b) the introduction of new performance indices that measure the probability of success over the entire frequency band, and (c) the introduction of explicit meta-models to provide sufficiently fast probability evaluations through Monte Carlo simulation. The key to the performance indices are limit-state functions formed at all discrete, contiguous, frequencies. Each limit-state function establishes a conformance region in terms of the random design variables. The probabilities of the conformance regions are correctly combined to provide a single series-system index to be maximized by adjusting distribution parameters. The simple explicit meta-model is based on Kriging and performance measures at arbitrary design sets are efficiently calculated. Error analysis suggests ways to predict and control the errors with regards to meta-model fitting and probability calculations and so the method appears sufficiently accurate for engineering applications. The proposed methodology has applications in numerous areas such as electrical filters and structural mechanics – all with n-dof and multiple responses. The Performance indices can be evaluated at any frequency over any number of frequency ranges. A case study of a vibration absorber mechanism shows how the new methodology provides an improved and timely design with controllable accuracy when compared with previous proposals that employed modal frequencies.     

Feature separation using ICA for a one-dimensional time series and its application in fault detection

Principal component analysis (PCA) is a method that transforms multiple data series into uncorrelated data series. Independent component analysis (ICA) is a method that separates multiple data series into independent data series. Both methods have been used in fault detection. However, both require signals from at least two separate sensors. To overcome this requirement and utilize the fault detection capability of ICA and PCA, we propose to use wavelet transform to pre-process the data collected from a single sensor and then use the coefficients of the wavelet transforms at different scales as input to ICA and PCA. The effectiveness of this method is demonstrated by applying it to both a simulated signal series and a vibration signal series collected from a gearbox. The results show that the method of combining wavelet transform and ICA works better than the method of combining wavelet transform and PCA for impulse detection based on a one-dimensional vibration data series.     

Acoustic response behaviour of panels mounted with equipment and its prediction using statistical energy analysis

Responses of non-uniform panels, like equipment panels of spacecraft, are not presently estimated using Statistical Energy Analysis. It is demonstrated that by treating the equipment as separate subsystems, with appropriate coupling loss factors for their connectivity, the response levels of the equipment can be estimated. The coupling loss factors are determined by the wave attenuation caused by the changes in structural properties introduced by the equipment. The estimated responses are valid at locations away from the boundary, in this case interface of the equipment. Information on the vibration levels at the interface of the equipment is necessary, especially to arrive at the random vibration loads of the equipment. A technique is developed to predict the vibration levels at locations at any distance from the interface of the equipment and validated by experiments. The prediction model is based on the interference pattern of the bending waves due to the reflection at the boundaries. It is seen that the existing techniques are suitable in estimating the responses of equipment only if the equipment behaves like mass attached at a point. But the technique developed in this study predicts the responses of the equipment that have large interface area.     

Sensitivity analysis of an auto-correlation-function-based damage index and its application in structural damage detection

Structural damage detection using time domain vibration responses has advantages such as simplicity in calculation and no requirement of a finite element model, which attracts more and more researchers in recent years. In present paper, a new approach to detect the damage based on the auto correlation function is proposed. The maximum values of the auto correlation function of the vibration response signals from different measurement points are formulated as a vector called Auto Correlation Function at Maximum Point Value Vector, AMV for short. The relative change of the normalized AMV before and after damage is used as the damage index to locate the damage. Sensitivity analysis of the normalized AMV with respect to the local stiffness shows that the normalized AMV has a sharp change around the local stiffness change location, which means the normalized AMV is a good indicator to detect the damage even when the damage is very small. Stiffness reduction detection of a 12-story frame structure is provided to illustrate the validity, effectiveness and the anti-noise ability of the proposed method. Comparison of the normalized AMV and the other correlation-function-based damage detection method shows the normalized AMV has a better detectability.     

Statistical damage identification of structures with frequency changes

Model updating methods based on structural vibration data have being rapidly developed and applied to detect structural damage in civil engineering. But uncertainties existing in the structural model and measured vibration data might lead to unreliable damage detection. In this paper a statistical damage identification algorithm based on frequency changes is developed to account for the effects of random noise in both the vibration data and finite element model. The structural stiffness parameters in the intact state and damaged state are, respectively, derived with a two-stage model updating process. The statistics of the parameters are estimated by the perturbation method and verified by Monte Carlo technique. The probability of damage existence is then estimated based on the probability density functions of the parameters in the two states. A higher probability statistically implies a more likelihood of damage occurrence. The presented technique is applied to detect damages in a numerical cantilever beam and a laboratory tested steel cantilever plate. The effects of using different number of modal frequencies, noise level and damage level on damage identification results are also discussed.     

Layout optimization methodology of piezoelectric transducers in energy-recycling semi-active vibration control systems

An optimization methodology is proposed for the piezoelectric transducer (PZT) layout of an energy-recycling semi-active vibration control (ERSAVC) system for a space structure composed of trusses. Based on numerical optimization techniques, we intend to generate optimal location of PZTs under the constraint for the total length of PZTs. The design variables are set as the length of the PZT on each truss based on the concept of the ground structure approach. The transient problems of the mechanical and electrical vibrations based on the ERSAVC theory are considered as the equations of state. The objective is to minimize the integration of the square of all displacement over the whole analysis time domain. The sensitivity of the objective function is derived based on the adjoint variable method. Based on these formulations, an optimization algorithm is constructed using the fourth-order Runge–Kutta method and the method of moving asymptotes. Numerical examples are provided to illustrate the validity and utility of the proposed methodology. Using the proposed methodology, the optimal location of PZTs for the vibration suppression for multi-modal vibration is studied, which can be benchmark results of further study in the context of ERSAVC systems.     

Effect of domain selection for compact representation of spatial variation of Head-Related Transfer Function in all directions based on Spatial Principal Components Analysis

In this article, compact representation of spatial variation of Head-Related Transfer Function (HRTF) or its corresponding inverse Fourier transform, namely Head-Related Impulse Response (HRIR) based on Principal Components Analysis (PCA), which is called the Spatial Principal Component Analysis (SPCA), is investigated, focusing on effect of domain selection. The SPCA was carried out for a database of HRTFs in all directions by selecting the domain as one of the HRIRs, the complex HRTFs, the frequency amplitudes of HRTFs, and log-amplitudes of HRTFs. In the latter two cases the minimum phase approximation was incorporated. Comparison of the accuracy in both time and frequency domains showed that the most compact representation is obtained by using the frequency amplitudes of HRTFs when the minimum phase approximation is acceptable, and the complex HRTFs bring about the most compact representation when the minimum phase approximation is not acceptable.     

Application of multi-scale chirplet path pursuit and fractional Fourier transform for gear fault detection in speed up and speed-down processes

Demodulation is very important for gear fault detection. However, the demodulation is substantially complicated by the non-stationary nature of the signal during the speed-up and speed-down processes. As such, we propose a new technique to detect gear faults under such conditions based on the multi-scale chirplet path pursuit (MSCPP) algorithm and the fractional Fourier transform (FrFT) method. With the MSCPP algorithm, the instantaneous frequency of the signal component with the largest energy in the multi-components gear vibration signal can be estimated. Then according to the estimated instantaneous frequency, the vibration signal segment whose instantaneous frequency curve approximated as either an ascending or descending linear segment can be obtained from the original gear vibration signal. In other words, the vibration signal segment that can be regarded as a multi-component linear frequency modulated (LFM) signal is extracted. As the FrFT is suitable for multi-component LFM signal analysis, it is then applied to demodulating this vibration signal segment and hence detecting local gear faults based on the revealed modulation phenomenon. The effectiveness of the proposed method has been demonstrated by both simulation and experimental data.     

SEA and contribution analysis for interior noise of a high speed train

This paper presents a detailed Statistical Energy Analysis (SEA) and contribution analysis of the interior noise of a high-speed train through extensive simulations and measurements. The SEA model was developed based on the actual geometrical parameters of a benchmark high-speed coach. Sound transmission loss levels of the structural components of the car body, which are required in the SEA model, were tested in a dedicated acoustic laboratory following international standard ISO 140-3:1995. Modal densities of these structural components were derived from measured frequency response functions using the modal counting method. Damping loss factors were obtained using the half-power bandwidth method and the vibration attenuation method. By considering the relationship between sound radiation and power transmission, coupling loss factors between structures and cavities were estimated. Source inputs to the SEA model were derived from field experiment data. Interior noise due to those sources was predicted using the SEA model and the prediction was generally in good agreement with measurement. Contribution analysis was then performed using this validated model through parametric study, and this analysis was further examined experimentally. In conclusion, for the coach that was investigated in this paper, the key factors for interior noise are sidewall vibration, bogie area noise, and floor sound transmission loss. Based on this and other engineering considerations, an interior noise control strategy can be defined.     

Structural health monitoring and damage assessment using a novel time series analysis methodology with sensor clustering

This study presents a novel time series analysis methodology to detect, locate, and estimate the extent of the structural changes (e.g. damage). In this methodology, ARX models (Auto-Regressive models with eXogenous input) are created for different sensor clusters by using the free response of the structure. The output of each sensor in a cluster is used as an input to the ARX model to predict the output of the reference channel of that sensor cluster. Two different approaches are used for extracting Damage Features (DFs) from these ARX models. For the first approach, the coefficients of the ARX models are directly used as the DFs. It is shown with a 4 dof numerical model that damage can be identified, located and quantified for simple models and noise free data. To consider the effects of the noise and model complexity, a second approach is presented based on using the ARX model fit ratios as the DFs. The second approach is first applied to the same 4 DOF numerical model and to the numerical data coming from an international benchmark study for noisy conditions. Then, the methodology is applied to the experimental data from a large scale laboratory model. It is shown that the second approach performs successfully for different damage cases to identify and locate the damage using numerical and experimental data. Furthermore, it is observed that the DF level is a good indicator for estimating the extent of the damage for these cases. The potential and advantages of the methodology are discussed along with the analysis results. The limitations of the methodology, recommendations, and future work are also addressed.     

On the tuning of variable piezoelectric transducer circuitry network for structural damage identification

The concept of using piezoelectric transducer circuitry with tunable inductance has been recently proposed to enhance the performance of frequency-shift-based damage identification method. While this approach has shown promising potential, a piezoelectric circuitry tuning methodology that can yield the optimal damage identification performance has not been synthesized. This research aims at advancing the state-of-the-art by exploring the characteristics of inductance tuning such that the enrichment of frequency measurements can be effectively realized to highlight the damage occurrence. Analysis shows that when the inductance is tuned to accomplish eigenvalue curve veering, the change of system eigenvalues induced by structural damage will vary significantly with respect to the change of inductance. Therefore, by tuning the inductance near the curve-veering range, one may obtain a family of frequency response functions that could effectively reflect the damage occurrence. When multiple tunable piezoelectric transducer circuitries are integrated to the mechanical structure, multiple eigenvalue curve veering can be simultaneously accomplished, and a series of inductance tunings can be formed by accomplishing curve veering between different pairs of system eigenvalues. It will then be shown that, to best characterize the damage occurrence, the favorable inductance tuning sequence should be selected as that leads to a “comprehensive” set of eigenvalue curve veering, i.e., all measurable natural frequencies undergo curve veering at least once. An iterative second-order perturbation-based algorithm is used to identify the locations and severities of the structural damages based on the frequency measurements before and after the damage occurrence. Numerical analyses on benchmark beam and plate structures have been carried out to examine the system performance. The effects of measurement noise on the effectiveness of the proposed damage identification method are also evaluated. It is demonstrated that the damage identification results can be significantly improved by using the variable piezoelectric transducer circuitry network with the favorable inductance-tuning scheme proposed in this research.     

Structural damage evolution assessment using the regularised time step integration method

This paper presents an approach to identify both the location and severity evolution of damage in engineering structures directly from measured dynamic response data. A relationship between the change in structural parameters such as stiffness caused by structural damage development and the measured dynamic response data such as accelerations is proposed, on the basis of the governing equations of motion for the original and damaged structural systems. Structural damage parameters associated with time are properly chosen to reflect both the location and severity development over time of damage in a structure. Basic equations are provided to solve the chosen time-dependent damage parameters, which are constructed by using the Newmark time step integration method without requiring a modal analysis procedure. The Tikhonov regularisation method incorporating the L-curve criterion for determining the regularisation parameter is then employed to reduce the influence of measurement errors in dynamic response data and then to produce stable solutions for structural damage parameters. Results for two numerical examples with various simulated damage scenarios show that the proposed method can accurately identify the locations of structural damage and correctly assess the evolution of damage severity from information on vibration measurements with uncertainties.